a neuroanatomical substrate linking perceptual stability ... · page 1 of 38 í î a...

Accepted manuscripts are peer-reviewed but have not been through the copyediting, formatting, or proofreadingprocess.

Copyright © 2019 Watanabe et al.This is an open-access article distributed under the terms of the Creative Commons Attribution 4.0 International license, whichpermits unrestricted use, distribution and reproduction in any medium provided that the original work is properly attributed.

This Accepted Manuscript has not been copyedited and formatted. The final version may differ from this version.

Research Articles: Behavioral/Cognitive

A neuroanatomical substrate linking perceptual stability to cognitiverigidity in autism

Takamitsu Watanabe1,2, Rebecca P Lawson3,4, Ylva S. E. Walldén3 and Geraint Rees1,3

1UCL Institute of Cognitive Neuroscience, 17 Queen Square, London, WC1N 3AZ, UK.2RIKEN Centre for Brain Science, 2-1 Hirosawa, Wako, Saitama, 351-0198, Japan.3Wellcome Centre for Human Neuroimaging, UCL, 12 Queen Square, London, WC1N 3BG, UK.4Department of Psychology, University of Cambridge, Downing St, Cambridge, CB2 3EB, UK.

https://doi.org/10.1523/JNEUROSCI.2831-18.2019

Received: 4 November 2018

Revised: 15 April 2019

Accepted: 21 May 2019

Published: 18 June 2019

Author contributions: T.W. and G.R. designed research; T.W., R.P.L., and Y.S.E.W. performed research; T.W.analyzed data; T.W. wrote the first draft of the paper; T.W., R.P.L., and G.R. edited the paper; T.W. and G.R.wrote the paper.

Conflict of Interest: The authors declare no competing financial interests.

This work was supported by a Marie-Curie Individual Fellowship from European Commission (656161; TW),YAMAHA Sports Challenge Fellowship (TW), Grant-in-aid for Research Activity Start-up (TW), Grant-in-aid forResearch Activity (TW), a Royal Society Wellcome Trust Henry Dale Fellowship (206691; RPL), and a WellcomeTrust Senior Clinical Research Fellowship (100227; GR). We thank all the participants who took part in thisexperiment. We also acknowledge Dr Paul Forbes for the management of participants.

Corresponding author: Takamitsu Watanabe, MD PhD, [email protected], UCL Institute ofCognitive Neuroscience, UK., RIKEN Centre for Brain Science, Japan

Cite as: J. Neurosci 2019; 10.1523/JNEUROSCI.2831-18.2019

Alerts: Sign up at www.jneurosci.org/alerts to receive customized email alerts when the fully formatted versionof this article is published.

of 38

1

A neuroanatomical substrate linking perceptual stability to cognitive rigidity in autism 2

3

Takamitsu Watanabe1,2, Rebecca P Lawson3,4 , Ylva S. E. Walldén3 , Geraint Rees1,3 4

5

1. UCL Institute of Cognitive Neuroscience, 17 Queen Square, London, WC1N 3AZ, UK. 6

2. RIKEN Centre for Brain Science, 2-1 Hirosawa, Wako, Saitama, 351-0198, Japan. 7

3. Wellcome Centre for Human Neuroimaging, UCL, 12 Queen Square, London, WC1N 3BG, UK. 8

4. Department of Psychology, University of Cambridge, Downing St, Cambridge, CB2 3EB, UK. 9

10

Corresponding author 11

Takamitsu Watanabe, MD PhD 12

[email protected] 13

UCL Institute of Cognitive Neuroscience, UK. 14

RIKEN Centre for Brain Science, Japan 15

16

Number of the display items: ten figures and five tables 17

18

Abbreviated title 19

Perceptual and cognitive rigidity in autism 20

21

Author contributions 22

TW and GR designed this study. TW conducted the behavioural experiments. RPL and YSEW 23

collected neuroimaging data. TW analysed the data. TW, RPL, YSEW, and GR wrote the manuscript. 24

of 38

Abstract 25

26

Overly stable visual perception seen in individuals with autism spectrum disorder (ASD) is related to 27

higher-order core symptoms of the condition. However, the neural basis by which these seemingly 28

different symptoms are simultaneously observed in individuals with ASD remains unclear. Here, we 29

aimed to identify such a neuroanatomical substrate linking perceptual stability to autistic cognitive 30

rigidity, a part of core restricted, repetitive behaviours (RRB). First, using a bistable visual 31

perception test, we measured the perceptual stability of 22 high-functioning adults with ASD and 22 32

age-/IQ-/sex-matched typically developing human individuals, and confirmed over-stable visual 33

perception in autism. Next, by employing a spontaneous task-switching test, we identified that the 34

individuals with ASD were more likely to repeat the same task voluntarily and spontaneously, and 35

such rigid task-switching behaviour was associated with the severity of their RRB symptoms. We 36

then compared these perceptual and cognitive behaviours and found a significant correlation between 37

them for individuals with ASD. Finally, we identified that this behavioural link was supported by a 38

smaller grey matter volume (GMV) of the posterior superior parietal lobule (pSPL) in individuals 39

with ASD. Moreover, this smaller GMV in the pSPL was also associated with the RRB symptoms 40

and replicated in two independent datasets. Our findings suggest that the pSPL could be one of the 41

neuroanatomical mediators of cognitive and perceptual inflexibility in autism, which could help a 42

unified biological understanding of the mechanisms underpinning diverse symptoms of this 43

developmental disorder. 44

45

Significance statement 46

47

Behavioural studies show perceptual over-stability in autism spectrum disorder (ASD). However, 48

neural mechanisms by which such sensory symptoms can coexist and often correlate with seemingly 49

separate core symptoms remain unknown. Here, we have identified such a key neuroanatomical 50

substrate. We have revealed that over-stable sensory perception of individuals with autism is linked 51

with their cognitive rigidity, a part of core restricted, repetitive behaviours symptoms, and such a 52

behavioural link is underpinned by a smaller grey matter volume in the posterior superior parietal 53

lobule in autism. These findings uncover a key neuroanatomical mediator of autistic perceptual and 54

cognitive inflexibility and would ignite future studies on how the core symptoms of autism interact 55

with its unique sensory perception. 56

of 38

Introduction 57

58

Studies of autism spectrum disorder (ASD) have mainly investigated two core aspects of the 59

condition: socio-communicational difficulties and restricted, repetitive behaviours (RRB) (American 60

Psychiatric Association, 2013). In addition to this trend, behavioural and clinical studies in the last 61

decade have accumulated evidence for connections between these core symptoms of autism and its 62

unique sensory perception (Robertson and Baron-Cohen, 2017): the core symptoms of ASD are 63

correlated with hypersensitivity (Boyd et al., 2009; Boyd et al., 2010), over-stability (Robertson et al., 64

2013; Freyberg et al., 2015; Robertson et al., 2016), and prediction error (Lawson et al., 2017) in 65

visual perception; also, the core symptoms are associated with over-evaluation of auditory 66

information (Lawson et al., 2017) and unique interpretation of smell (Endevelt-Shapira et al., 2018). 67

68

However, little is known about the neural substrates underpinning the coexistence of such altered 69

sensory perception with more complex, higher-order core symptoms in individuals with ASD 70

(Robertson and Baron-Cohen, 2017). Neuroimaging studies identified larger signal variability in the 71

visual, auditory, and somatosensory brain areas of individuals with ASD (Dinstein et al., 2012; 72

Haigh et al., 2015), and the disruption of GABAergic inhibitory neuronal systems in their early 73

visual cortex (Robertson et al., 2016). However, these studies did not report associations between the 74

neural responses seen in autism and behaviours related to its sensory/core symptoms. Another study 75

found that ASD core symptoms are correlated with the size of population visual receptive fields in 76

the primary visual cortex (Schwarzkopf et al., 2014), but this work did not identify relationships 77

between the neural architecture in autism and its sensory symptoms. Although a recent population-78

based genetics study found genes that were commonly related to core autistic traits and unique 79

perceptual sensitivity in typically developing (TD) individuals (Taylor et al., 2018), this research 80

neither directly examined these genetic observations in cohorts with ASD nor investigated neural 81

mechanisms involving the overlapping genetic patterns. Currently, therefore, the neurobiological 82

bases that link sensory symptoms to core symptoms in autism remain unclear (Robertson and Baron-83

Cohen, 2017). 84

85

Here, we aimed to identify such a biological link by focusing on overly stable visual perception 86

(Robertson et al., 2013; Freyberg et al., 2015; Robertson et al., 2016) and cognitive rigidity, an 87

aspect of the RRB symptoms (Lopez et al., 2005; D'Cruz et al., 2013; Dajani and Uddin, 2015). We 88

chose these perceptual and cognitive tendencies in ASD because our previous studies imply that both 89

of them could be supported by overly stable brain dynamics in autism: although it has not been 90

of 38

confirmed in ASD, perceptual stability during the bistable perception was predicted by less frequent 91

transitions of large-scale brain activity patterns in TD individuals (Watanabe et al., 2014a); likewise, 92

the general cognitive skills of high-functioning adults with ASD were explained by overly stable 93

transitory brain dynamics (Watanabe and Rees, 2017). Based on these observations, we hypothesised 94

that the stable visual perception in ASD and its rigid cognitive styles should share neuroanatomical 95

substrates. 96

97

We examined this hypothesis by searching for neuroanatomical features of cortical grey matter that 98

were correlated with both the perceptual stability and the cognitive rigidity in high-functioning adults 99

with ASD. 100

101

Materials and Methods 102

103

Overall study design 104

105

First, we behaviourally compared the stability of visual perception and cognitive rigidity. 106

107

The stability of visual perception was measured in a test of bistable visual perception with a 108

structure-from-motion stimulus (Kanai et al., 2010; Kanai et al., 2011; Watanabe et al., 2014a; 109

Megumi et al., 2015)(Fig. 1A). We used this stimulus because it contains no explicit social/biological 110

information, which enabled us to evaluate perception with the least effects of the social symptoms of 111

ASD. 112

113

Cognitive rigidity was evaluated by observing spontaneous task-switching (TS) between two 114

cognitive tasks (shape and brightness tasks; Fig. 1B). As in previous studies (Arrington and Logan, 115

2004; Poljac et al., 2012), for each trial, participants were asked to freely choose one of the two tasks 116

and to conduct the task accurately and quickly (Fig.1C). We counted how many times participants 117

repeated the same task and used the task-repetition length as an index of cognitive rigidity. 118

119

As a control, the participants also performed instructed TS tests, in which they were given explicit 120

instructions indicating which of the two tasks they had to conduct on each trial (Fig. 1D). The order 121

of instructions (i.e., the task order) was determined by the record of that participant’s behaviour in 122

the previous spontaneous TS tests, so that the participants had seemingly the same task-switching 123

experience in both the instructed and spontaneous TS tests (Fig. 1E). This experimental paradigm 124

of 38

allowed us to examine whether the cognitive rigidity seen in the spontaneous TS test could be simply 125

explained by basic cognitive/sensorimotor skills seen in the instructed TS test. 126

127

Using these behavioural data, we first quantified perceptual stability in ASD (Fig. 2) and the basic 128

skills for these task-switching behaviours (Fig. 3). Then, we examined the randomness and 129

spontaneity of participants’ responses during the spontaneous TS tests (Fig. 4) and investigated 130

associations between perceptual and cognitive inflexibility (Fig. 5). 131

132

By combining these behavioural observations with structural MRI data, we conducted a voxel-based 133

morphometry (VBM) and searched for brain areas whose cortical grey matter volumes (GMVs) were 134

commonly correlated with perceptual stability and with cognitive rigidity in individuals with autism 135

(Figs. 6-8) and controls (Fig. 9). 136

137

Finally, to confirm the associations between the GMVs of the focal brain regions and cognitive 138

rigidity in ASD, we also compared the GMVs and the severity of the RRB symptoms in our dataset 139

and two independent neuroimaging datasets (Di Martino et al., 2014) (Fig. 10). 140

141

142 Fig. 1. Experimental design. 143

of 38

A. In a bistable visual perception test, participants were presented with a structure-from-144

motion stimulus to both their eyes and asked to report when their perception switched 145

between downward and upward rotating with a button press. 146

B. Task-switching (TS) tests consisted of shape and brightness tasks. On each trial, 147

participants had to choose the circle (shape task) or the brightest figure (brightness task). 148

C. In the spontaneous TS test, participants could freely choose which task to perform for each 149

trial. We quantified cognitive rigidity as how long participants repeated the same task 150

continuously. 151

D. In the instructed TS test, participants were given instructions about which task to perform 152

for each trial. 153

E. The order of the instruction in the instructed TS tests was determined based on their own 154

responses during the spontaneous TS tests. Therefore, the participants performed the same 155

task-switching pattern in the two TS tests. 156

157

Participants 158

159

We recruited 27 high-functioning right-handed adults with ASD and 27 age-/IQ-/sex-matched TD 160

individuals. This sample size was determined based on previous studies that examined perceptual 161

stability in autism (Robertson et al., 2013; Robertson et al., 2016). Participants were diagnosed by 162

independent clinicians according to DSM-IV or ICD-10, and the severity of their core symptoms was 163

assessed by a qualified administrator using ADOS (Lord et al., 1989). To reduce the heterogeneity of 164

the participants, we focused on high-functioning adults with ASD (Full IQ/Verbal IQ/Practice IQ ≥ 165

85). The IQ was assessed by the Wechsler Adult Intelligence Scale (third edition, UK). We also 166

excluded individuals with any other neuropsychiatric disorders, those whose were taking psychiatric 167

medicines at the time of the experiments, and those who could not undergo all the experiments. In 168

total, five ASD and five TD individuals were excluded for incompletion of the experiments due to 169

technical problems or their visibly noisy MRI data probably due to head motion. We analysed 170

behavioural and MRI data of the remaining 22 ASD (3 females) and 22 TD (4 females) individuals 171

(Table 1). 172

173

This study was approved by the UCL ethics committee. All the participants provided written 174

informed consent and were financially compensated for their participation and travel expenses. 175

176

Table 1. Demographic data

of 38

TD ASD P value Number of participants 22 22 - Age (mean ± s.e.m.) 30.8±1.6 33±2.0 0.41 Sex Female: 4 Female: 3 - Laterality Right-handed Right-handed Full IQ (mean ± s.e.m.) 112.8±3.0 119.7±2.6 0.1 Verbal IQ (mean ± s.e.m.) 114.5±2.9 120.6±3.1 0.1 Performance IQ (mean ± s.e.m.) 108.0±3.2 114.7±2.6 0.18 ADOS Social (mean ± s.e.m.) - 2.7±0.3 - ADOS Comm (mean ± s.e.m.) - 6.2±0.5 - ADOS RRB (mean ± s.e.m.) - 0.8±0.2 - 177

178

Test of bistable visual perception 179

180

In the test of bistable visual perception, the participants were presented with a structure-from-motion 181

stimulus (Fig. 1A), a sphere consisting of 200 moving white dots in a black background (Watanabe 182

et al., 2014a). The dots moved sinusoidally upwards and downwards (angular velocity, 120°/sec) 183

with a fixation cross (0.1°×0.1°) at the centre of the 21.5-inch LCD monitor (Samsung Sync Master 184

2233, resolution: 1680×1050). In each run, the participants were asked to see this stimulus for 90 sec 185

with their chins put on a chin rest. They were instructed to push one of the three buttons according to 186

their visual perception: one for upward rotation, another for downward rotation, and the other for 187

unsure/mixture perception. After sufficient training sessions, the participants repeated this run five 188

times. We conducted the stimulus presentation and response recording with PsychToolbox 3 in 189

MATLAB (MathWorks, Inc). 190

191

Consistent with our previous studies (Watanabe et al., 2014a; Megumi et al., 2015), mixture 192

perception was rare in both participant groups (1.7±0.3% of all stimulus presentation times, 193

mean±s.d.). Therefore, this study focused on the time during which participants had clear awareness 194

about the direction of the rotation. For each participant, we measured the duration of such clear 195

perception, and calculated the median of the duration to evaluate their perceptual stability. We used 196

the ‘median’ duration because perceptual durations showed long-tailed distributions in both 197

participant groups (Fig. 2A). 198

199

Spontaneous task-switching test 200

201

of 38

In the spontaneous TS test (Arrington and Logan, 2004; Poljac et al., 2012), participants were 202

presented with visual stimuli consisting of four figures with different shapes and brightness (Fig. 1B), 203

and were asked to perform a shape task or a brightness task. In the shape task, they had to identify a 204

specific shape (here, a circle), whereas they should identify the brightest figure in the brightness task. 205

They were asked to press one of four buttons to indicate which figure in the display they chose for 206

each task as accurately and quickly as possible. If they did not press any button in 3 sec, the trial 207

automatically ended, and the next trial started. 208

209

Participants practiced the two types of task separately by repeating a 30-sec trial until they could 210

respond correctly (≥95% of accuracy) and quickly (reaction time ≤2 sec). Then, the participants 211

underwent the spontaneous TS test consisting of five 3-min runs. 212

213

For each trial, they could freely choose which task they would conduct. As in previous studies 214

(Arrington and Logan, 2004; Poljac et al., 2012), we asked them, “You have to choose which task to 215

perform on each trial. Ideally, you should perform each task randomly but on about half of the trials. 216

Sometimes you will repeat the same task and sometimes you will switch from one to another. We 217

don’t want you to count the number of times you’ve done each task or alternate strictly between tasks 218

to make it sure you do each one half the time. Just try to do them randomly.” 219

220

Based on the chosen figures, we retrospectively inferred the selectedd tasks, and classified each trial 221

into a shape or brightness trial. Participants rarely chose a figure that could not be classified into 222

either a shape or brightness task (the proportion of such unclassifiable trials ≤1.5%; Fig. 4E). 223

Therefore, we excluded such unclassifiable trials for the following analysis. 224

225

Instructed task-switching test 226

227

In the instructed TS test, participants were given clear instructions about which task to choose. In 228

each trial, a letter appeared at the centre of the screen, and indicated the instruction. For example, ‘C’ 229

meant a shape task in which the participants should choose a circle, whereas ‘B’ meant a brightness 230

task (Fig. 1D). Participants were asked to perform a task according to this instruction accurately and 231

quickly. 232

233

The sequence of the instruction letters was determined based on the behaviours in the spontaneous 234

TS test that the participants underwent right before (Fig. 1E). The unclassifiable trials in the 235

of 38

spontaneous TS task were simply omitted in the instructed TS task. Thus, the participants were 236

supposed to have seemingly the same task-repetition/switching experience as in the spontaneous TS 237

test. 238

239

Behavioural analyses for task-switching tests 240

241

For both the TS tests, we calculated the mean reaction time (RT) for task repetition, mean RT for 242

task switching, and switch cost (mean RT for tasks switching – mean RT for task repetition) for each 243

participant. The response accuracy was also estimated in the instructed TS test, and the median of the 244

task-repetition length was calculated in the spontaneous TS test. The mean values were used to 245

represent the RT because RT showed normal distributions (P ≥ 0.24 in Shapiro-Wilk test), whereas 246

the median values were chosen as a representative metric for the task-repetition length because they 247

showed long-tailed distributions (Fig. 4E). All these behavioural indices were first estimated for each 248

experiment, and then averaged across the five runs. 249

250

We used behavioural data during the instructed TS test to evaluate sensorimotor skills to perform 251

task-switching. We compared the RT for task repetition, RT for task switch, switch cost, and 252

response accuracy between the ASD and TD groups. Then, to examine the spontaneity in the 253

spontaneous TS test, we compared the RT for the task repetition, RT for the task switch, and switch 254

cost between the instructed and spontaneous TS tests. 255

256

Validation of the randomness in the spontaneous TS test 257

258

To examine the randomness in the spontaneous TS test, we compared the task-switching/repetition 259

behaviours in the TS test with the state-switching/duration patterns in a simple random-walk model. 260

261

First, we conducted a one-dimensional Markov-chain random-walk simulation (Fig. 4G). The 262

randomness of this simulation was iterated by changing the transition probability, PTrans, between 0.5 263

and 0.9 by 0.1 (PTrans = 0.5: the most random movements, PTrans = 0.9: the most systematic and 264

regular movements). The number of the hidden states between the extreme states (state hi in Fig. 4G) 265

was also iterated between one and five. For each parameter set, we simulated this random walk for 266

105 steps and counted the state duration by measuring the number of the steps taken to move from 267

one extreme to the other (e.g., movements from the state A to the state B in Fig. 4G). Then, we 268

calculated the distribution of the state duration and compared it with that of the task-repetition length 269

of 38

in the spontaneous TS test. KL divergence was used to quantify the difference between the two 270

distribution. Note that, to directly compare the shape of the distributions, we normalised the 271

distributions before assessing KL divergence. 272

273

This normalisation procedure also lessened a potentially confounding effect that would be induced 274

by the following constraint in the random-walk simulation. For simplicity, we did not consider the 275

possibility of self-recurrent movements in the random walk. This constraint would allow continuous 276

staying in a hidden state over time, and, theoretically, the state duration in this random walk would 277

be shorter compared to the otherwise case. However, such reduction in the state duration should be 278

nearly cancelled out by the normalisation of the resultant distribution. Thus, this random-walk 279

constraint is considered to have no significant effects on the calculation of KL divergence. 280

281

Finally, for each participant group, we examined associations between the KL divergence and the 282

randomness of the random walk (i.e., PTrans), and identified which PTrans could produce the most 283

similar state-duration distribution compared to that of the task-repetition length. 284

285

Comparison of cognitive rigidity 286

287

After confirming the behavioural randomness and spontaneity in the spontaneous TS test, we 288

compared the median task-repetition length in the test between the ASD and TD groups. 289

290

Note that the task-repetition length was sufficiently short compared to the length of one run. In every 291

3-min run, both the ASD and TD participants completed ≥200 trials, whereas the median task-292

repetition length was around three trials in both the groups and the ASD-TD gap was less than one 293

trial (Fig. 5A). Therefore, even if the ASD individuals showed longer task-repetition length, they 294

experienced task switching as often (~65 times per run, ~330 times per participant) as TD individuals 295

did (~68 times per run, ~340 times per participant). Given this, we could assume that the longer task-296

switching length in the ASD group did not reduce the size of the behavioural data sampling, and, 297

thus, did not affect the accuracy of the statistical analyses using the behavioural records of the ASD 298

individuals. 299

300

Behavioural associations 301

302

of 38

In each participant group, we calculated the Pearson correlation between the percept duration in the 303

bistable perception test and the task-repetition length in the spontaneous TS test. 304

305

We also estimated associations between these behavioural metrics and the clinical scores. For the 306

social symptoms, we calculated correlation coefficients between the experimental indices and ADOS 307

social+comm scores. For RRB symptoms, we compared the behavioural indices between individuals 308

with different ADOS RRB scores using an analysis of variance (ANOVA). 309

310

Neuroimaging experiment and analysis 311

312

We obtained T1-weighted brain images using a 3T MRI scanner (Trio, Siemens Medical Systems) 313

with the 32-channel head coil at the Wellcome Trust Centre for Neuroimaging at UCL (TR 7.92ms, 314

TE 2.48ms, Flip angle 16º, spatial resolution 1mm cubic). These MRI images were pre-processed in 315

SPM12 in the following steps. First, the images were segmented into grey matter, white matter, and 316

cerebrospinal fluid in native space by the New Segment Toolbox (Ashburner and Friston, 2005). 317

Then, the segmented grey matter images were aligned, warped to a template space (ICBM space 318

template for European), resampled down to 1.5mm isotropic voxels, and registered to a participant-319

specific template by the DARTEL Toolbox (Ashburner, 2007). Using deformation parameters 320

calculated by the DARTEL toolbox, the grey matter images were normalised to MNI space and 321

smoothed with a Gaussian kernel (FWHM = 8mm). We set FWHM at 8mm because the value was 322

widely used (see a review by (Nickl-Jockschat et al., 2012)) and previous studies demonstrated that 323

this size of spatial smoothing would improve the accuracy of VBM (Pepe et al., 2014) without 324

increasing false positive rates (Scarpazza et al., 2015). This preprocessing using DARTEL Toolbox 325

included a modulation process to preserve the volume of a particular tissue within a voxel (Good et 326

al., 2001). Thus, signal values in the preprocessed images should indicate GMVs. 327

328

We analysed these preprocessed MRI images of the ASD group with a multiple regression model in 329

SPM12. In VBM, we separately searched for brain regions whose GMVs were correlated with [1] 330

the median percept duration and [2] the median task-repetition length. Both the regression analyses 331

used individual ages and FIQ scores as covariates. We then identified brain regions whose GMVs 332

were positively/negatively correlated with each behavioural index in the ASD data (PFDR = 0.05). 333

The anatomical labels for the brain regions were determined based on the Automated Anatomical 334

Labelling atlas (Tzourio-Mazoyer et al., 2002) and Harvard-Oxford Atlas. 335

336

of 38

Next, we conducted a conjunction analysis (Nichols et al., 2005) to specify brain regions whose 337

GMVs were correlated with both the behavioural indices. Technically, we searched for the overlaps 338

between the two type of statistical brain map calculated above. The statistical threshold of this 339

conjunction analysis was equivalent or more stringent compared to those in previous studies 340

adopting this analysis approach (Chadick and Gazzaley, 2011; Watanabe et al., 2014b). 341

342

We then investigated whether the GMV of the overlapping region (here, posterior superior parietal 343

lobule, pSPL) was significantly different between the TD and ASD groups. 344

345

Brain-behaviour associations 346

347

We conducted a partial correlation analysis to examine the associations between the GMV of the 348

pSPL, perceptual stability, cognitive rigidity, and RRB symptoms. 349

350

We then performed a non-parametric mediation analysis to test whether the pSPL was a mediator 351

linking the perceptual over-stability to cognitive rigidity in autism. The GMV of the pSPL was set as 352

a mediator variable, whereas the median percept duration and median task-repetition length were 353

used as an independent and dependent variable, respectively. We also conducted a whole-brain 354

voxel-wise non-parametric mediation analysis using the same independent/dependent variables (PFDR 355

< 0.05). 356

357

Finally, we performed a structural equation modelling (SEM) analysis to examine whether the GMV 358

of the pSPL was related to domain-general behavioural/mental flexibility. The model had a latent 359

variable representing a domain-general flexibility and was fitted to the behavioural and GMV data 360

with a maximum likelihood method. As in a previous study (Schermelleh-Engel et al., 2003), 361

goodness of fit was evaluated based on adjusted goodness of fit index (AGFI), comparative fit index 362

(CFI), root mean square error of approximation (RMSEA), and standardized root mean square 363

residual (SRMR). 364

365

To validate results of the above neuroanatomical analysis, we repeated the same VBM and brain-366

behaviour analyses using data collected from only male individuals with ASD. Also, we repeated the 367

neuroanatomical analyses using data of the TD individuals. 368

369

Reproducibility of GMV alteration in autism 370

of 38

371

We examined the reproducibility of the smaller GMV of the right pSPL using two independent MRI 372

datasets (the University of Utah and ETH Zürich) shared in ABIDE and ABIDE II (Di Martino et al., 373

2014) (Table 2). The participants were selected with the same criteria as in the original analysis, and 374

their T1-weighted brain images were pre-processed in the same manner as stated above. 375

376

We extracted the GMV of the right pSPL from each participant using the grey-matter mask identified 377

in the original analysis (a yellow area in Fig. 6B). We then compared the GMV of the pSPL between 378

the ASD and TD groups, and examined whether the individuals with ASD showing less GMV in the 379

pSPL had severer RRB symptoms. 380

381

To examine the specificity of the pSPL, we conducted whole-brain voxel-wise search for brain 382

regions whose GMVs were significantly smaller in the ASD group compared to the controls (PFDR < 383

0.05). 384

385

For every brain region found this VBM, we then examined associations between its GMV and the 386

severity of RRB symptoms. First, we calculated the average GMV for each significant cluster for 387

each individual with ASD. Next, we conducted two-sample t-tests of the mean GMV between the 388

individuals with ASD with larger RRB scores (ADOS RRB score ≥ 1) and the other individuals with 389

ASD (i.e., ADOS RRB score = 0). 390

391 392 Table 2. Demographic data for reproducibility tests 393 TD ASD P value University of Utah Number of participants 26 24 – Age (Mean ± SD) 25.3±6.3 25.3±5.5 0.96 Sex Male Male – Handedness Right Right Full IQ 112.6±12.0 109.9±14.2 0.47 Verbal IQ 112.1±12.1 107.5±14.2 0.22 Performance IQ 110.3±10.4 110.25±16.0 0.98 ADOS Social – 4.8±2.1 – ADOS Communication – 7.0±2.3 – ADOS RRB – 1.1±1.1 –

ETH Zürich Number of participants 15 10 –

of 38

Age (Mean ± SD) 23.2±3.3 21.5±2.7 0.2 Sex Male Male – Handedness Right Right – Full IQ 114.7±9.2 108.9±12.8 0.21 Verbal IQ 115±12.1 109.5±14.5 0.23 Performance IQ 112.7±10.1 107.0±11.9 0.21 ADOS Social – 6.7±1.3 – ADOS Communication – 2.5±1.1 – ADOS RRB – 0.5±0.97 – 394

Statistics 395

396

All post-hoc t-tests were corrected using Bonferroni correction. We set α = 0.05/3 when comparing 397

three behavioural indices between the two TS tests (Figs. 4A-4C), α = 0.05/2 when examining 398

correlations between the task-repetition length and two types of ADOS score (Fig. 5C) and when 399

evaluating associations between the GMV of pSPL and two behavioural indices (Fig. 6C). When 400

comparing GMV in the four different brain regions between the ASD and TD groups, we set α = 401

0.05/4 (Fig. 6G). In the whole-brain neuroimaging analyses, FDR correction was adopted. 402

Correlations were evaluated as Spearman’s coefficients when either of them was non-parametric 403

(e.g., ADOS score). Otherwise, they were calculated as Pearson correlation coefficients. 404

405

Results 406

407

Perceptual stability 408

409

First, using the bistable perception test (Fig. 1A), we found a significantly longer percept duration in 410

the ASD group compared to the controls (t42 = 3.0, P = 0.0049 in a two-sample t-test; Figs. 2A and 411

2B). This perceptual stability in autism was consistently observed throughout the test (F1,210 = 7.6, P 412

= 0.0068 in a repeated-measures two-way ANOVA with a [ASD/TD] ⨉ [five runs] structure; Fig. 413

2C). 414

415

of 38

416Fig. 2. Behaviour in the bistable visual perception test. 417

Both the ASD and TD groups showed a skewed distribution of percept duration in a bistable 418

perception test (panel A). The median duration was longer in the ASD group (panel B), which 419

was observed throughout the experiment (panel C). Error bars, s.e.m. 420

421

Sensorimotor skill for task switching 422

423

Next, we quantified cognitive rigidity of autism by comparing behaviours between the instructed and 424

spontaneous task-switching (TS) tests (Figs. 1B-1E). 425

426

Behaviours during the instruction-based TS test were not different between the ASD and TD groups. 427

Both the groups accurately performed the two tasks (response accuracy > 93.5%; no group difference, 428

t42 = 0.02, P = 0.98 in a two-sample t-test; Fig. 3A) and showed equivalent reaction time (RT) for 429

task repetition and for task switching (t42 < 0.83, P ≥ 0.41 in two-sample t-tests; Fig. 3B). Thus, the 430

switch cost (RT for task switching – RT for task repetition) was almost the same between the two 431

groups (t42 = 0.56, P = 0.58 in a two-sample t-test; Fig. 3C). This seemingly intact task-switching 432

ability of the individuals with ASD did not change throughout the test (F1,210 = 0.041, P = 0.84 in a 433

repeated-measures two-way ANOVA; Fig. 3D). 434

435

In addition, the switch cost — a widely used index for cognitive flexibility in TD-participant studies 436

(Aron et al., 2004; Dajani and Uddin, 2015; Watanabe et al., 2015b) — did not show significant 437

differences between individuals with different severity of RRB symptoms (F2,19 = 1.4, P = 0.27 in a 438

one-way ANOVA; Fig. 3E). 439

440

These results suggest that high-functioning adults with ASD had sufficient sensorimotor skills to 441

accurately perform the two tasks and smoothly switch between them according to the instructions. 442

443

of 38

444 Fig. 3. Behaviour in instructed task-switching (TS) test. 445

In the instructed TS test, both groups showed similar behaviour. The ASD and TD groups 446

responded accurately (panel A) with almost the same reaction time (RT) (panel B). The switch 447

cost was significantly larger than zero in both the groups (P < 10–5 in one-sample t-tests; *** in 448

panel C) with no significant difference between the cohorts (P = 0.58 in a two-sample t-test; 449

panel C) throughout the test (panel D). The switch cost of the individuals with ASD was not 450

related to the severity of their RRB symptoms (P = 0.27 for the main effect of ADOS RRB 451

scores in a one-way ANOVA; panel E). Error bars, s.e.m. 452

453

Spontaneous and random task switching 454

455

Participants experienced the same task-switching sequence in the spontaneous and instructed TS tests, 456

because the timing of the switches in the instructed TS test was yoked to previous per-participant 457

spontaneous switch timings (Fig. 1E). However, behavioural responses during the spontaneous TS 458

test were distinct from those during the instructed TS test mainly in the following two points (Fig. 4), 459

each of which indicates the spontaneity and randomness of the task-switching behaviour, 460

respectively. 461

462

First, the spontaneity was supported by reaction time analyses. Compared to the instructed TS test, 463

the ASD and TD participants repeated tasks in the spontaneous TS test more quickly (F1,84 = 8.6, P = 464

of 38

0.004 in a repeated-measures two-way ANOVA with a [ASD/TD] ⨉ [Spontaneous/Instructed] 465

structure; Fig. 4A) and switched them more smoothly (F1,84 = 15.0, P = 0.0002 in a repeated-466

measures two-way ANOVA; Fig. 4B) with significantly less switch cost (F1,84 = 21.0, P < 0.0001; 467

Fig. 4C). These behaviours are reasonable if the participants switched and repeated the tasks 468

spontaneously. 469

470

Second, the randomness was supported by the task-switching frequencies and intervals. The 471

participants performed the two tasks for almost equal frequencies (Fig. 4D) with a small proportion 472

of unclassified trials (mean ≤ 1.5%; Fig. 4E), which is reasonable if the participants switched tasks 473

randomly. Moreover, the task-repetition length showed a skewed distribution with a long tail (Fig. 474

4F), which was similar to the distribution of state-switching intervals in a random walk (Fig. 4G). 475

That is, when a one-dimensional Markov-chain random walk was determined by more random 476

transitions (e.g., PTrans = 0.5 in Fig. 4G), the resultant distribution of the state-switching intervals was 477

more similar to the actual task-repetition-length distribution seen in this spontaneous TS test (Figs. 478

4H and 4I). 479

480

These observations suggest that the participants performed the spontaneous TS test as spontaneously 481

and randomly as possible. 482

483

of 38

484Fig. 4. Validation of spontaneous task-switching test.485

A-C. In the spontaneous TS test, ASD and TD participants repeated the tasks more quickly 486

(panel A) and switched them more smoothly (panel B) than in the instructed TS test. The 487

switch cost was not significantly different from zero and significantly less than in the 488

instructed TS test (panel C). These results suggest that the participants repeated and 489

switched the tasks spontaneously. *, PBonferroni < 0.05. 490

D. The ASD and TD participants performed the two types of task at almost equal frequencies. 491

Error bars, s.e.m. 492

E. The mean proportion of the unclassified trials in the spontaneous TS test was less than 493

1.5% throughout the test and not different between the ASD and TD groups. Error bars, s.d. 494

of 38

F-I. In both the participant groups, the task-repetition length showed a skewed, long-tailed 495

distribution (panel F). To evaluate the randomness of the task switch, we compared the 496

distribution with that seen in one-dimensional Markov-chain random-walk simulations 497

(panel G). When the simulation adopted more random transition probability (e.g., PTrans = 498

0.5), the difference between the distributions (KL divergence) became smaller (panel H). 499

The lines in the panel H show the mean KL divergence across different numbers of the 500

hidden states varying between 1 and 5 shown in the panel I. Error bars, s.e.m. 501

502

Cognitive rigidity 503

504

Such spontaneous TS behaviours enabled us to detect cognitive rigidity in ASD. The ASD and TD 505

groups showed equivalent RTs for task repetition, RTs for task switching, and the switch cost (P > 506

0.55 in two-sample t-tests; Figs. 4A-4C), but the task-repetition length was significantly longer in the 507

ASD group (t42 = 4.5, P < 0.0001, PBonferroni < 0.05 in a two-sample t-test; Fig. 5A). This was 508

consistently observed throughout the test (F1,210 = 8.6, P = 0.0038 in a repeated-measures two-way 509

ANOVA; Fig. 5B). 510

511

The prolonged task-repetition length in the ASD group could not be explained by more basic 512

cognitive/sensorimotor skills: the task-repetition length was not significantly correlated with the RT 513

for task repetition, the RT for task switching, or the switch cost in either type of TS test (|r| < 0.19, P 514

> 0.40); neither the IQ scores (FIQ, PIQ, VIQ) nor the age of the participants were associated with 515

this cognitive rigidity index (|r| < 0.21, P > 0.34). 516

517

Instead, this longer task repetition in the ASD group was associated with the severity of the RRB 518

symptoms (F2,19 = 6.2, P = 0.009 for the main effect of ADOS RRB scores in a one-way ANOVA; 519

Fig. 5C), but was not correlated with the severity of socio-communicational symptoms (Spearman’s 520

rho = 0.13, P = 0.56; Fig. 5C). 521

522

Taken together, the behavioural results from the two TS tests indicate that prolonged task-repetition 523

length in individuals with ASD is not a consequence of their sensorimotor skills or basic cognitive 524

abilities, but is related to the RRB symptoms. This observation supports our assumption that, at least 525

in high-functioning individuals with ASD, task-repetition lengths in the spontaneous TS test are 526

associated with their cognitive rigidity. 527

528

of 38

Associations between perceptual stability and cognitive rigidity 529

530

We then assessed links between perceptual stability and cognitive rigidity by comparing the percept 531

duration in the bistable perception test and the task-repetition length in the spontaneous TS test. We 532

found that, in the individuals with ASD, both behavioural indices were deviated from those of TD 533

individuals to almost the same degree (F1,84 = 0.82, P > 0.37 in a repeated-measures two-way 534

ANOVA; Fig. 5D) and were significantly correlated with each other (r = 0.45, P = 0.035; Fig. 5E). 535

536

537

538 Fig. 5. Associations between perceptual stability and cognitive rigidity. 539

A-C. The task-repetition length in the spontaneous TS test was significantly larger in the 540

ASD group (panel A), which was consistently seen throughout the test (panel B). In 541

the individuals with ASD, the task-repetition length was specifically related to the 542

severity of the RRB symptoms (panel C). In the panels a and c, the Y axis shows the 543

average of the median task-repetition length across the five experimental runs for each 544

participant. In the panel B, the Y axis represents the mean of the median task-545

repetition length across participants. 546

of 38

D and E. We compared the perceptual stability with the cognitive rigidity in the ASD group. 547

The deviations from the neurotypical responses were almost the same in the two type 548

of behavioural indices (panel D), and these two indices were correlated with each 549

other in the ASD group (panel E). ‘d’ indicates Cohen’s effect size. Error bars, s.e.m. 550

551

Neuroanatomical substrates for perceptual and cognitive inflexibility 552

553

We then conducted a voxel-based morphometry (VBM) and searched for a neuroanatomical 554

substrate underlying this behavioural link between the perceptual stability and cognitive rigidity in 555

autism (Table 3). 556

557

Table 3. Results of VBM using ASD data. 558

Laterality Peak coordinate (MNI)

Region X Y Z t value Correlation with percept duration Positive

IOC Right 40 –64 0 4.6 Negative

pSPL/Angular gyrus Right 36 –54 46 5.1

Correlation with task-repetition length Positive

ACC Right 8 22 18 4.5 MFG Right 46 10 40 4.4

Negative pSPL/sLOC Right 38 –60 42 4.8 IOC, Inferior occipital cortex; pSPL, posterior superior parietal lobule; ACC, Anterior Cingulate 559 Cortex; MFG, Middle frontal gyrus; sLOC, superior lateral occipital cortex. Anatomical labels were 560 determined based on the AAL atlas. 561 562

First, we found that perceptual stability in the individuals with ASD was positively correlated with 563

the grey matter volume (GMV) of the inferior occipital cortex (IOC) (t = 4.6, PFDR-corrected < 0.05; red 564

in Fig. 6A), and negatively correlated with the GMV of the posterior superior parietal lobule (pSPL) 565

(t = 5.1, PFDR-corrected < 0.05; red in Fig. 6B). 566

567

In contrast, the cognitive rigidity in autism showed positive correlations with the GMVs of the 568

anterior cingulate cortex (ACC) and middle frontal gyrus (MFG) (t ≥ 4.4, PFDR < 0.05; red in Fig. 569

6A), and a negative correlation with the GMV of the pSPL (t = 4.8, PFDR < 0.05; red in Fig. 6B). 570

of 38

571

That is, the pSPL was the only region whose GMV was significantly associated with both perceptual 572

stability and cognitive rigidity in autism. In fact, the GMV in the overlapping pSPL area (a yellow 573

area in Fig. 6B) was correlated with both perceptual and cognitive rigidity in the ASD group (r ≤ –574

0.62, P ≤ 0.003, PBonferroni < 0.05; Fig. 6C). In addition, the GMV of the pSPL was smaller when the 575

individuals with ASD had severer RRB symptoms (t20 = 3.1, P = 0.0062 in a two-sample t-test; Fig. 576

6D). 577

578

These brain-behaviour associations seen in the pSPL were not explained by differences in whole-579

brain GMV. The whole-brain GMV in the ASD group was marginally larger compared to the TD 580

group (t42 = 2.1, P = 0.048 in a two-sample t-test; Fig. 6E), but was not correlated with either 581

perceptual stability or cognitive rigidity (|r| ≤ 0.1, P > 0.64). Moreover, even the relative GMV of the 582

pSPL — a ratio of the pSPL GMV to the whole-brain GMV — was still associated with both the 583

behavioural indices (r ≤ –0.48, P ≤ 0.01; Fig. 6F). 584

585

We then compared GMVs of brain regions found in the above VBM (i.e., IOC, pSPL, ACC, and 586

MFG) between the ASD and TD groups. Despite logical independence between this GMV 587

comparison and the above VBM (Figs. 6A and 6B), we found that the pSPL was, again, the only 588

region with a significant difference in the GMV (t42 = 4.2, Puncorrected = 0.0012, PBonferroni < 0.05 in 589

two-sample t-tests; Fig. 6G). This difference survived even when we considered effects of whole-590

brain GMV (t42 = 9.6, P < 10–5 in a two-sample t-test; Fig. 6H). 591

592

Taken together, these findings suggest that the perceptual over-stability and cognitive rigidity in 593

autism share some neural bases and the diminished GMV of the pSPL constitutes such brain 594

mechanisms. 595

596

of 38

597 598

Fig. 6. Neuroanatomical results. 599

A and B. Red areas indicate voxels whose GMVs were significantly correlated with the median 600

percept duration in the bistable perception test, whereas blue areas show those whose 601

GMVs were significantly correlated with the median task-repetition length in the 602

spontaneous TS test. The yellow area was an overlap between the red and blue regions 603

(centre, [36, –53, 44] in MNI coordinates, pSPL). For presentation purpose, the 604

statistic brain maps adopted t = 3.0 as their thresholds. IOC, inferior occipital cortex. 605

ACC, anterior cingulate cortex. MFG, middle frontal gyrus. pSPL, posterior superior 606

parietal lobule. See also Table 3 for details. 607

of 38

C. The GMV of the overlapping pSPL region was negatively correlated with both the 608

percept duration and task-repetition length. 609

D. The GMV of the overlapping pSPL was significantly smaller in ASD individuals with 610

severer RRB symptoms. 611

E. The whole-brain GMV was marginally larger in the ASD group. Error bars, s.e.m. 612

F. The relative GMV of the pSPL — a ratio of the GMV of the pSPL to the whole-brain 613

GMV — still shows significant negative correlations with the median percept duration 614

in the bistable perception test (left panel) and the median task-repetition length in the 615

spontaneous TS test (right panel). Each circle represents each individual with ASD. 616

G. Among the four regions, the pSPL was the only region whose GMV was significantly 617

different between the ASD and TD groups. Error bars, s.e.m. *, PBonferroni < 0.05. 618

H. The relative GMV of the pSPL was significantly smaller in the individuals with ASD 619

compared to the controls. Error bars, s.e.m. 620

621

Associations between the pSPL, perceptual stability and cognitive rigidity 622

623

Based on the above findings, we then examined whether the pSPL was a key neural substrate linking 624

perceptual over-stability to cognitive rigidity in autism. 625

626

This notion was consistent with results of a partial correlation analysis (Fig. 7A). The correlation 627

between the percept duration and the task-repetition length (Fig. 5E) was explainable by the 628

associations between the GMV of the pSPL and these two autistic behaviours. 629

630

A non-parametric mediation analysis provided more direct evidence. This analysis showed that the 631

GMV of the pSPL had a significantly large indirect effect ( × = 0.54, P = 0.009; Fig. 7B), which 632

suggests that the pSPL would be a mediator between the perceptual and cognitive rigidity. In 633

addition, a whole-brain non-parametric mediation analysis demonstrated that a region in the pSPL 634

was the only brain area that had a significant indirect effect ( × = 0.62, Puncorrected = 0.0003, PFDR < 635

0.05; Fig. 7C). 636

637

Furthermore, a structural equation modelling (SEM) analysis indicates that the pSPL area could be 638

related to a neural mechanism supporting domain-general behavioural/mental flexibility (Fig. 7D). 639

640

of 38

These findings suggest that the pSPL is one of the key brain regions linking the perceptual stability 641

to the cognitive inflexibility in autism and the smaller GMV of the pSPL would result in the two 642

seemingly different symptoms of autism. These observations were preserved even when we 643

conducted the analyses using only the data of the males with ASD (Fig. 8). 644

645

646 Fig. 7. Brain-behaviour associations 647

A. A partial correlation analysis implies that the smaller GMV of the pSPL in autism 648

linked the perceptual over-stability to the cognitive rigidity and, consequently, RRB 649

symptoms. 650

B. A non-parametric mediation analysis suggests that the smaller GMV of the pSPL 651

would be a mediator linking perceptual stability to cognitive rigidity. This analysis 652

used the median percept duration, GMV of the pSPL, and median task-repetition 653

length as an independent variable, mediator variable, and dependent variable, 654

respectively. indicates a regression coefficient of the percept duration on the GMV 655

of the pSPL, and denotes that of the GMV of the pSPL on the task-repetition length. 656 ’ represents the direct effect of the percept duration on the task-repetition length. 657 × indicates the indirect effect of the percept duration on the task-repetition length 658

via the pSPL. 659

of 38

C. Coloured clusters are brain regions whose GMVs had significant indirect effects 660

( × ) in a whole-brain non-parametric mediation analysis with the same independent 661

and dependent variables as in the panel B. For presentation purpose, the statistic brain 662

map adopted P = 0.01 as its threshold. A significant cluster was found in the pSPL (P 663

= 0.0003 in [34, –52, 44] in MNI coordinates). 664

D. A structural equation modelling (SEM) analysis, in which domain-general flexibility 665

was introduced as a latent variable, indicates that the pSPL is related to domain-666

general behavioural/mental flexibility and that the smaller GMV of the pSPL in 667

autism would attenuate such flexibility and induce perceptual over-stability and 668

cognitive rigidity. Note that the arrows did not indicate causal relationships between 669

the variables. λ, path coefficients; e and , residual terms; AGFI, adjusted goodness of 670

fit index; CFI, comparative fit index; RMSEA, root mean square error of 671

approximation; SRMR, standardised root mean square residual. 672

673

674 Fig. 8. Neuroanatomical results based on male data. 675

We repeated the neuroanatomical analyses using only the data recorded from the males with 676

ASD (N = 19) and the male controls (N = 18). 677

of 38

A and B. In the VBM, we found almost the same pSPL region (a yellow area in panel A) 678

whose GMV was inversely correlated with both the percept duration and task-679

repetition length (panel B). 680

C. The GMV of this pSPL was larger in the ASD individuals with smaller RRB 681

scores compared to those with larger RRB score. 682

D. The GMV of the pSPL area was smaller in the ASD group than in the TD 683

group. 684

E. A non-parametric mediation analysis showed that the pSPL is a mediator 685

linking the percept duration to the task-repetition length. For the details of the 686

analysis and abbreviations, see the legend for Fig. 7B. 687

F. An SEM analysis indicates that the pSPL area could be related to domain-688

general behavioural/mental flexibility. For the details of the analysis and 689

abbreviations, see the legend for Fig. 7D. 690

691

Neuroanatomical results in TD individuals 692

693

We then examined whether these brain-behaviour associations observed in the pSPL were specific to 694

individuals with ASD or also seen in the TD individuals. 695

696

First, we directly tested this question: we calculated the GMV-behaviour correlations by applying the 697

pSPL area determined in the above analyses of the ASD data (i.e., a yellow area in Fig. 6B) to the 698

TD dataset as an anatomical mask. We found that, even in the TD dataset, the GMV of this pSPL 699

area was inversely correlated with both the percept duration (r = –0.42, P = 0.04) and the task-700

repetition length (r = –0.49, P = 0.015). 701

702

Second, we conducted a more thorough examination by repeating the entire VBM with the TD 703

dataset only. Compared to the observations based on the ASD data (Table 3, Figs. 6A and 6B), 704

similar brain regions were associated to the perceptual stability and/or cognitive rigidity (Table 4, 705

Figs. 9A and 9B; PFDR < 0.05), and, again, the pSPL was the only region whose GMV was 706

commonly correlated with the two behavioural indices. 707

708

In fact, the GMV of the overlapping pSPL area (a yellow region in Fig. 9B) showed significant 709

negative correlations with both the percept duration (r = –0.50, P = 0.01) and the task-repetition 710

length (r = –0.54, P = 0.006). Moreover, as seen in the analysis of the ASD data, a non-parametric 711

of 38

mediation analysis indicates that the pSPL area would be a mediator linking the perceptual stability 712

to the cognitive rigidity ( × = 0.50, P = 0.009; Fig. 9C). This pSPL area was also found in a 713

whole-brain non-parametric mediation analysis (PFDR < 0.05; Fig. 9D). Furthermore, an SEM 714

analysis implies that the pSPL could be related to domain-general behavioural/mental flexibility (Fig. 715

9E). 716

717

These results suggest that the associations between the pSPL and the perceptual/cognitive rigidity are 718

not specific to autism and the pSPL could be one of the essential brain regions supporting human 719

behavioural/mental flexibility. 720

721

722 Fig. 9. Neuroanatomical results based on TD data. 723

A and B. The GMVs in the red areas were correlated with the median percept duration in the 724

bistable perception test, whereas those in the blue areas were correlated with the 725

median task-repetition length in the spontaneous TS test. The yellow area was an 726

overlap between the red and blue clusters. The overlapping area (panel B) was 727

located around [34, –46, 40] in MNI coordinates. For presentation purpose, the 728

statistic brain maps adopted t = 3.0 as their thresholds. IOC, inferior occipital 729

cortex. aSPL/pSPL, anterior/posterior superior parietal lobule. MFG, middle frontal 730

gyrus. SFG, superior frontal gyrus. See also Table 4 for details. 731

of 38

C. A non-parametric mediation analysis suggests that, even in TD individuals, the 732

pSPL is a mediator linking perceptual stability to cognitive rigidity. For the details 733

of the analysis and abbreviations, see the legend for Fig. 7B. 734

D. Coloured clusters are brain regions whose GMVs had significant indirect effects 735

( × ) in a whole-brain non-parametric mediation analysis. For presentation 736

purpose, the statistic brain map adopted P = 0.01 as its threshold. A significant 737

cluster was found in the pSPL (P = 0.0008 in [36, –52, 38] in MNI coordinates) 738

E. An SEM analysis indicates that the pSPL could be related to domain-general 739

behavioural/mental flexibility even in TD individuals. For the details of the 740

analysis and abbreviations, see the legend for Fig. 7D. 741

742

Table 4. Results of VBM analysis using TD data 743

Laterality Peak coordinate (MNI)

Region X Y Z t value Correlation with duration in bistable perception test Positive

aSPL Right 30 –40 48 4.7 IOC Right 32 –82 10 4.3

Negative pSPL Right 30 –62 52 4.9

Correlation with task-repetition length in spontaneous TS test Positive

MFG Right 28 38 40 4.4 MFG Right 42 14 34 5.1

Negative SFG Right 18 40 34 4.7 Precuneus Left –8 –37 56 4.8 Precuneus Right 8 –60 66 4.5

pSPL Right 42 –62 36 4.6 aSPL, anterior superior parietal lobule; IOC, Inferior occipital cortex; pSPL, posterior superior 744 parietal lobule; MFG, Middle frontal gyrus; SFG, superior frontal gyrus. Anatomical labels were 745 determined based on the AAL atlas. 746 747

Replication of smaller GMV of pSPL in autism 748

749

Finally, we examined reproducibility for the smaller GMV of the pSPL in autism and its association 750

with the RRB symptoms. This test was conducted using two independent neuroimaging datasets in 751

ABIDE (Di Martino et al., 2014) (datasets recorded in University of Utah and ETH Zürich; Table 2). 752

of 38

753

In both datasets, the GMV of the pSPL (a yellow area in Fig. 6B) was significantly smaller in high-754

functioning adults with ASD compared to demographically-matched TD individuals (P ≤ 0.003 in 755

two-sample t-tests; Fig. 10A). Within the ASD groups, the GMV of the pSPL was decreased when 756

the individuals with ASD had larger ADOS RRB scores (P ≤ 0.04 in two-sample t-tests; Fig. 10B), 757

which is consistent with the observation shown in Fig. 6D. 758

759

In addition, this smaller GMV and association with the RRB symptoms were specific to the pSPL. In 760

the both datasets, voxel-wise neuroanatomical comparisons between the TD and ASD groups 761

identified significant decreases in the GMV in other brain regions, such as right hippocampus and 762

amygdala, in autism (PFDR < 0.05; Table 5); however, none of them showed significant associations 763

with the severity of the RRB symptoms (t ≤ 0.84, P ≥ 0.41 in two-sample t-tests; right two columns 764

in Table 5). Only in the pSPL, the GMV was smaller in the ASD individuals with larger RRB scores 765

(RRB ≥ 1) compared to those with smaller ones (RRB = 0) (t ≥ 2.7, P ≤ 0.022 in two-sample t-tests). 766

767

These results add evidence for associations between the diminished GMV in the pSPL and the 768

cognitive inflexibility in autism. 769

770

771 Fig. 10. Reproducibility of smaller GMV of pSPL in autism 772

We examined the reproducibility of the diminished GMV seen in the pSPL in autism 773

and its link with the RRB symptoms (Figs. 6D and 6G) using two independent 774

neuroimaging datasets of high-functioning adults with ASD and age-/IQ-/sex-matched 775

controls (University of Utah and ETH Zürich; Table 2). In both datasets, the GMV of 776

the pSPL (a yellow area in Fig. 6B) was decreased in the ASD individuals (panel A) and 777

of 38

showed smaller values when the individuals with ASD had severer RRB symptoms 778

(panel B). 779

780

Table 5. Brain regions whose GMVs were smaller in ASD and associated with RRB 781 symptoms 782

Right / Left

Peak coordinate (MNI)

ASD v TD

Association with RRB symptoms (RRB=0 v RRB≥1)

Region X Y Z t value t value P value Dataset collected in University of Utah Amygdala Right 30 –2 –20 4.7 0.64 0.53 ACC Right 8 44 8 4.6 –0.84 0.41 Anterior Insula Right 36 22 0 4.2 0.75 0.46 pSPL Right 32 –54 50 4.6 2.7 0.014 Dataset collected in ETH Zürich Hippocampus Right 26 –4 –22 4.2 0.39 0.71 Anterior Insula Right 36 14 –10 3.9 0.83 0.43 pSPL Right 30 –52 58 4.1 2.8 0.022 ACC, anterior cingulate cortex. pSPL, posterior superior parietal lobule. RRB, repetitive restricted 783 behaviour. T values and P values in “RRB=0 v RRB≥1” indicate results of two-sample t-tests of 784 GMVs between the ASD individuals with zero of RRB scores and those with ≥1 of RRB scores. 785 Positive t values denote that the GMVs were larger in individuals with zero of RRB, whereas the 786 negative values indicate that the GMVs were smaller. 787 788

Discussion 789

790

This case-control study has shown that overly stable visual perception of high-functioning adults 791

with ASD is linked to their cognitive rigidity, and this behavioural link is associated with the smaller 792

GMV of the pSPL. We first found that over-stable sensory perception seen in the bistable perception 793

test in individuals with ASD is correlated with cognitive rigidity measured by the spontaneous task-794

switching test. Then, we identified the smaller GMV in the pSPL as a neuroanatomical substrate 795

filling the gap between these seemingly separate behaviours in individuals with ASD. 796

797

The current behavioural findings can be seen as evidence for a link between over-stable visual 798

perception of high-functioning adults with ASD and (a part of) their RRB symptoms. Multiple 799

previous studies reported behavioural associations between autistic sensory perception and ASD core 800

symptoms (Robertson et al., 2012; Robertson et al., 2013; Lawson et al., 2017; Endevelt-Shapira et 801

al., 2018); however, the core symptoms examined in these earlier researches were socio-802

communicational symptoms or total ASD severity, and not RRB symptoms. Some behavioural 803

of 38

studies reported an association between hyper-/hypo-sensitivity in visual perception and RRB in 804

children with autism using fine clinical scales for the core symptoms such as Repetitive Behaviour 805

Scales-Revised (Boyd et al., 2009; Boyd et al., 2010), but the relationship has not been demonstrated 806

in an adult population. As a consequence, unlike hyper-/hypo-sensitivity, perceptual inflexibility has 807

not been yet included in a subcategory of the RRB symptoms even in the latest diagnosis criteria for 808

ASD (American Psychiatric Association, 2013). 809

810

This may be due to difficulty in finely quantifying the non-social features of autism in adults with 811

ASD. In clinical settings, the RRB symptoms of adults with autism are measured on a relatively 812

coarse scale. For example, in ADOS (Lord et al., 1989), the severity of RRB is ranked using a 813

limited range of integers (e.g., 0, 1, or 2 in this study and 0, 1, 2, or 3 in the ABIDE datasets used 814

here; Table 1 and Table 2). ADOS has been repeatedly validated as a tool to aid diagnosis (Lord et 815

al., 1989) and index severity (Gotham et al., 2009; Watanabe et al., 2015a) of this autism, but such a 816

relatively sparse scoring system of the RRB symptoms raises the possibility that this scale could not 817

capture nuanced individual differences in cognitive rigidity. Even in behavioural experiments, 818

autistic cognitive rigidity is known to be difficult to detect (Schmitz et al., 2006; Geurts et al., 2009; 819

Poljac and Bekkering, 2012; Yerys et al., 2015). In particular, high-functioning adults with ASD 820

often easily adapt themselves to instruction-based TS tests, and occasionally outperform neurotypical 821

individuals (Poljac et al., 2010; Poljac and Bekkering, 2012). In fact, the current instructed TS test 822

found no significant behavioural differences between the ASD and TD groups (Figs. 3A-3D). 823

824

In contrast, the spontaneous TS test detected cognitive rigidity — a part of the RRB symptoms 825

(Lopez et al., 2005; D'Cruz et al., 2013; Dajani and Uddin, 2015) — in the high-functioning adults 826

with ASD with a relatively large effect size (Cohen’s d = 1.4; Figs. 5A and 5D). The task-repetition 827

length recorded in this TS test provided more detailed information about individual differences in 828

cognitive rigidity and was related to the severity of the RRB symptoms (Fig. 5C). In addition, given 829

that previous studies also reported unique behavioural patterns in autistic people using a similar test 830

(Arrington and Logan, 2004; Poljac et al., 2012), this psychological paradigm appears to have a 831

robust sensitivity to detect cognitive rigidity in autism. This improved detectability may support our 832

ability to identify neuroanatomical bases related to cognitive rigidity in autism. 833

834

In the neuroanatomical investigation, we found that the smaller GMV in the pSPL was associated 835

with both autistic stable perception and cognitive rigidity. In terms of the relationship between the 836

brain area and bistable visual perception, this finding is consistent with previous reports about 837

of 38

dissociable functions of the anterior and posterior SPLs. A series of behavioural, neuroimaging, and 838

brain-stimulation studies show that the posterior SPL destabilises visual perception when viewing the 839

structure-from-motion stimulus, whereas the anterior SPL stabilises it (Kanai et al., 2010; Kanai et 840

al., 2011; Watanabe et al., 2014a; Megumi et al., 2015). Therefore, the smaller GMV in pSPL should 841

stabilise visual perception, which is consistent with the current observation. The findings in this work 842

strengthen our knowledge about the functional anatomy of the SPL and its role in bistable perception. 843

844

The association between the GMV of the pSPL and cognitive rigidity is also consistent with previous 845

reports about neural mechanisms underlying human cognitive flexibility (Cole and Schneider, 2007; 846

Cole et al., 2013; Watanabe and Rees, 2017). The frontoparietal network, including this region, is 847

considered essential for flexible coordination of different types of cognitive skill (Cole and 848

Schneider, 2007; Cole et al., 2013), and several studies reported associations between autistic 849

cognitive inflexibility and atypical neural activity of this brain network (Misic et al., 2015; Uddin et 850

al., 2015; Watanabe and Rees, 2017). In addition, a meta-analysis demonstrated a critical role of the 851

SPL in a wide range of task-switching behaviours (Kim et al., 2012), and another neuroimaging 852

study using similar voluntary task-switching test also suggested the importance of SPL in a voluntary 853

action (Poljac and Yeung, 2014). These previous findings are consistent with the results of the 854

current SEM analysis (Fig. 6G), in which the pSPL appears to be involved in domain-general mental 855

flexibility. 856

857

This study found diminished GMV in the pSPL in high-functioning adults with ASD in three 858

independent datasets (Figs. 6G and 10A, Table 5), which are consistent with previous findings on the 859

neuroanatomy of autism. Although we have to carefully consider effects of the heterogeneity of ASD 860

(Hardan et al., 2006; Salmond et al., 2007; Yu et al., 2011; Pua et al., 2017), multiple meta-analyses 861

reported significant decreases in the GMVs of the SPL (Via et al., 2011; Duerden et al., 2012) and 862

neighbouring parietal regions (Cauda et al., 2011; Via et al., 2011; Yu et al., 2011) in individuals 863

with ASD. The current findings can be seen as further biological evidence implying the critical role 864

of the pSPL in autism but now linking this finding to specific behaviours. 865

866

One of the major limitations of this study is that we did not examine functional brain architecture 867

underlying the autistic perceptual and cognitive inflexibility. Methodologically, this work focused on 868

establishing of a psychological paradigm to detect a link between perceptual and cognitive 869

inflexibility in autism. Based on this behavioural observation, we then searched for preliminary 870

neuroanatomical evidence for the behavioural link. Future research should examine how the pSPL 871

of 38

interacts (or fails to interact) with other brain areas during spontaneous task switching in individuals 872

with ASD. 873

874

Another limitation is in the relatively small sample size employed for the main experiment. The 875

sample size was determined by previous behavioural studies on perceptual stability in autism where 876

the number of participants was sufficient to reject the null hypothesis (Robertson et al., 2013; 877

Robertson et al., 2016). No previous study provided data about how strongly the spontaneous TS test 878

can detect the cognitive rigidity in high-functioning individuals with ASD, but the effect sizes 879

reported here provide a principle basis for determining future sample sizes necessary to identify 880

neural bases for such cognitive rigidity. 881

882

Also, we have to note that it remains unknown how the atypical neuroanatomical development of the 883

SPL in autism (Watanabe and Rees, 2016) is related to the development of cognitive flexibility. In 884

addition, our findings are not necessarily directly applicable to other sensory symptoms in autism. 885

Seemingly lower-level sensory symptoms, such as hyper/hypo-sensitivity, might not involve the SPL 886

area. For example, typical auditory perception might be underpinned by atypical neural architectures 887

in the superior temporal sulcus. At the same time, the SPL may also play a key role in these sensory 888

symptoms because the area is known to integrate a wide range of sensory information (Akrami et al., 889

2018) and be involved in the top-down allocation of attention over primary sensory regions (e.g., see 890

a review in (Moore and Zirnsak, 2017)). The smaller GMV in adults with ASD reported here may 891

affect how low-level sensory inputs are constrained by top-down expectations (Lawson et al., 2014; 892

Lawson et al., 2015; Palmer et al., 2017). Future studies will be necessary to clarify whether 893

feedback processing from the SPL to sensory regions is associated with perceptual sensitivity. 894

895

This case-control study has found a behavioural association between overly stable visual perception 896

and cognitive rigidity in high-functioning autistic adults. Moreover, we have identified the pSPL as 897

one of the major neuroanatomical bases supporting this behavioural association. These findings 898

should lead to future studies on how the ASD core symptoms interact with perceptual difficulties 899

experienced by individuals with autism, which will help us to comprehensively understand the 900

cognitive and behavioural styles of this disorder. 901

902

Acknowledgements 903

904

of 38

This work was supported by a Marie-Curie Individual Fellowship from European Commission 905

(656161; TW), YAMAHA Sports Challenge Fellowship (TW), Grant-in-aid for Research Activity 906

Start-up (TW), Grant-in-aid for Research Activity (TW), a Royal Society Wellcome Trust Henry 907

Dale Fellowship (206691; RPL), and a Wellcome Trust Senior Clinical Research Fellowship 908

(100227; GR). We thank all the participants who took part in this experiment. We also acknowledge 909

Dr Paul Forbes for the management of participants. 910

911

Competing interests 912

913

The authors declare no competing financial interests. 914

915

References 916

Akrami A, Kopec CD, Diamond ME, Brody CD (2018) Posterior parietal cortex represents sensory 917 history and mediates its effects on behaviour. Nature 554:368-372. 918

American Psychiatric Association D-TF (2013) Diagnostic and statistical manual of mental 919 disorders : DSM-5, 5th Edition. Washington, D.C.: American Psychiatric Association. 920

Aron AR, Monsell S, Sahakian BJ, Robbins TW (2004) A componential analysis of task-switching 921 deficits associated with lesions of left and right frontal cortex. Brain 127:1561-1573. 922

Arrington CM, Logan GD (2004) The cost of a voluntary task switch. Psychol Sci 15:610-615. 923 Ashburner J (2007) A fast diffeomorphic image registration algorithm. Neuroimage 38:95-113. 924 Ashburner J, Friston KJ (2005) Unified segmentation. Neuroimage 26:839-851. 925 Boyd BA, McBee M, Holtzclaw T, Baranek GT, Bodfish JW (2009) Relationships among Repetitive 926

Behaviors, Sensory Features, and Executive Functions in High Functioning Autism. Res 927 Autism Spectr Disord 3:959-966. 928

Boyd BA, Baranek GT, Sideris J, Poe MD, Watson LR, Patten E, Miller H (2010) Sensory features 929 and repetitive behaviors in children with autism and developmental delays. Autism Res 3:78-930 87. 931

Cauda F, Geda E, Sacco K, D'Agata F, Duca S, Geminiani G, Keller R (2011) Grey matter 932 abnormality in autism spectrum disorder: an activation likelihood estimation meta-analysis 933 study. J Neurol Neurosurg Psychiatry 82:1304-1313. 934

Chadick JZ, Gazzaley A (2011) Differential coupling of visual cortex with default or frontal-parietal 935 network based on goals. Nat Neurosci 14:830-832. 936

Cole MW, Schneider W (2007) The cognitive control network: Integrated cortical regions with 937 dissociable functions. Neuroimage 37:343-360. 938

Cole MW, Reynolds JR, Power JD, Repovs G, Anticevic A, Braver TS (2013) Multi-task 939 connectivity reveals flexible hubs for adaptive task control. Nat Neurosci 16:1348-1355. 940

D'Cruz AM, Ragozzino ME, Mosconi MW, Shrestha S, Cook EH, Sweeney JA (2013) Reduced 941 behavioral flexibility in autism spectrum disorders. Neuropsychology 27:152-160. 942

Dajani DR, Uddin LQ (2015) Demystifying cognitive flexibility: Implications for clinical and 943 developmental neuroscience. Trends Neurosci 38:571-578. 944

Di Martino A et al. (2014) The autism brain imaging data exchange: towards a large-scale evaluation 945 of the intrinsic brain architecture in autism. Mol Psychiatry 19:659-667. 946

Dinstein I, Heeger DJ, Lorenzi L, Minshew NJ, Malach R, Behrmann M (2012) Unreliable evoked 947 responses in autism. Neuron 75:981-991. 948

of 38

Duerden EG, Mak-Fan KM, Taylor MJ, Roberts SW (2012) Regional differences in grey and white 949 matter in children and adults with autism spectrum disorders: an activation likelihood 950 estimate (ALE) meta-analysis. Autism Res 5:49-66. 951

Endevelt-Shapira Y, Perl O, Ravia A, Amir D, Eisen A, Bezalel V, Rozenkrantz L, Mishor E, 952 Pinchover L, Soroka T, Honigstein D, Sobel N (2018) Altered responses to social 953 chemosignals in autism spectrum disorder. Nat Neurosci 21:111-119. 954

Freyberg J, Robertson CE, Baron-Cohen S (2015) Reduced perceptual exclusivity during object and 955 grating rivalry in autism. J Vis 15:11. 956

Geurts HM, Corbett B, Solomon M (2009) The paradox of cognitive flexibility in autism. Trends 957 Cogn Sci 13:74-82. 958

Good CD, Johnsrude IS, Ashburner J, Henson RN, Friston KJ, Frackowiak RS (2001) A voxel-based 959 morphometric study of ageing in 465 normal adult human brains. Neuroimage 14:21-36. 960

Gotham K, Pickles A, Lord C (2009) Standardizing ADOS scores for a measure of severity in autism 961 spectrum disorders. J Autism Dev Disord 39:693-705. 962

Haigh SM, Heeger DJ, Dinstein I, Minshew N, Behrmann M (2015) Cortical variability in the 963 sensory-evoked response in autism. J Autism Dev Disord 45:1176-1190. 964

Hardan AY, Keshavan MS, Sreedhar S, Vemulapalli M, Minshew NJ (2006) An MRI study of minor 965 physical anomalies in autism. J Autism Dev Disord 36:607-611. 966

Kanai R, Bahrami B, Rees G (2010) Human parietal cortex structure predicts individual differences 967 in perceptual rivalry. Curr Biol 20:1626-1630. 968

Kanai R, Carmel D, Bahrami B, Rees G (2011) Structural and functional fractionation of right 969 superior parietal cortex in bistable perception. Curr Biol 21:R106-107. 970

Kim C, Cilles SE, Johnson NF, Gold BT (2012) Domain general and domain preferential brain 971 regions associated with different types of task switching: a meta-analysis. Hum Brain Mapp 972 33:130-142. 973

Lawson RP, Rees G, Friston KJ (2014) An aberrant precision account of autism. Front Hum 974 Neurosci 8:302. 975

Lawson RP, Friston KJ, Rees G (2015) A more precise look at context in autism. Proc Natl Acad Sci 976 U S A 112:E5226. 977

Lawson RP, Mathys C, Rees G (2017) Adults with autism overestimate the volatility of the sensory 978 environment. Nat Neurosci 20:1293-1299. 979

Lopez BR, Lincoln AJ, Ozonoff S, Lai Z (2005) Examining the relationship between executive 980 functions and restricted, repetitive symptoms of Autistic Disorder. J Autism Dev Disord 981 35:445-460. 982

Lord C, Rutter M, Goode S, Heemsbergen J, Jordan H, Mawhood L, Schopler E (1989) Autism 983 diagnostic observation schedule: a standardized observation of communicative and social 984 behavior. J Autism Dev Disord 19:185-212. 985

Megumi F, Bahrami B, Kanai R, Rees G (2015) Brain activity dynamics in human parietal regions 986 during spontaneous switches in bistable perception. Neuroimage 107:190-197. 987

Misic B, Doesburg SM, Fatima Z, Vidal J, Vakorin VA, Taylor MJ, McIntosh AR (2015) 988 Coordinated Information Generation and Mental Flexibility: Large-Scale Network Disruption 989 in Children with Autism. Cereb Cortex 25:2815-2827. 990

Moore T, Zirnsak M (2017) Neural Mechanisms of Selective Visual Attention. Annu Rev Psychol 991 68:47-72. 992

Nichols T, Brett M, Andersson J, Wager T, Poline JB (2005) Valid conjunction inference with the 993 minimum statistic. Neuroimage 25:653-660. 994

Nickl-Jockschat T, Habel U, Michel TM, Manning J, Laird AR, Fox PT, Schneider F, Eickhoff SB 995 (2012) Brain structure anomalies in autism spectrum disorder--a meta-analysis of VBM 996 studies using anatomic likelihood estimation. Hum Brain Mapp 33:1470-1489. 997

of 38

Palmer CJ, Lawson RP, Hohwy J (2017) Bayesian approaches to autism: Towards volatility, action, 998 and behavior. Psychol Bull 143:521-542. 999

Pepe A, Dinov I, Tohka J (2014) An automatic framework for quantitative validation of voxel based 1000 morphometry measures of anatomical brain asymmetry. Neuroimage 100:444-459. 1001

Poljac E, Bekkering H (2012) A review of intentional and cognitive control in autism. Front Psychol 1002 3:436. 1003

Poljac E, Yeung N (2014) Dissociable neural correlates of intention and action preparation in 1004 voluntary task switching. Cereb Cortex 24:465-478. 1005

Poljac E, Poljac E, Yeung N (2012) Cognitive control of intentions for voluntary actions in 1006 individuals with a high level of autistic traits. J Autism Dev Disord 42:2523-2533. 1007

Poljac E, Simon S, Ringlever L, Kalcik D, Groen WB, Buitelaar JK, Bekkering H (2010) Impaired 1008 task switching performance in children with dyslexia but not in children with autism. Q J Exp 1009 Psychol (Hove) 63:401-416. 1010

Pua EPK, Bowden SC, Seal ML (2017) Autism spectrum disorders: Neuroimaging findings from 1011 systematic reviews. Research in Autism Spectrum Disorders 34:28-33. 1012

Robertson CE, Baron-Cohen S (2017) Sensory perception in autism. Nat Rev Neurosci 18:671-684. 1013 Robertson CE, Ratai EM, Kanwisher N (2016) Reduced GABAergic Action in the Autistic Brain. 1014

Curr Biol 26:80-85. 1015 Robertson CE, Martin A, Baker CI, Baron-Cohen S (2012) Atypical integration of motion signals in 1016

Autism Spectrum Conditions. PLoS One 7:e48173. 1017 Robertson CE, Kravitz DJ, Freyberg J, Baron-Cohen S, Baker CI (2013) Slower rate of binocular 1018

rivalry in autism. J Neurosci 33:16983-16991. 1019 Salmond CH, Vargha-Khadem F, Gadian DG, de Haan M, Baldeweg T (2007) Heterogeneity in the 1020

patterns of neural abnormality in autistic spectrum disorders: evidence from ERP and MRI. 1021 Cortex 43:686-699. 1022

Scarpazza C, Tognin S, Frisciata S, Sartori G, Mechelli A (2015) False positive rates in Voxel-based 1023 Morphometry studies of the human brain: should we be worried? Neurosci Biobehav Rev 1024 52:49-55. 1025

Schermelleh-Engel K, Moosbrugger H, Müller H (2003) Evaluating the fit of structural equation 1026 models: tests of significance and descriptive goodness-of-fit measures. Metho Psychol Res 1027 8:52. 1028

Schmitz N, Rubia K, Daly E, Smith A, Williams S, Murphy DG (2006) Neural correlates of 1029 executive function in autistic spectrum disorders. Biol Psychiatry 59:7-16. 1030

Schwarzkopf DS, Anderson EJ, de Haas B, White SJ, Rees G (2014) Larger extrastriate population 1031 receptive fields in autism spectrum disorders. J Neurosci 34:2713-2724. 1032

Taylor MJ, Gustafsson P, Larsson H, Gillberg C, Lundstrom S, Lichstenstein P (2018) Examining 1033 the Association Between Autistic Traits and Atypical Sensory Reactivity: A Twin Study. J 1034 Am Acad Child Adolesc Psychiatry 57:96-102. 1035

Tzourio-Mazoyer N, Landeau B, Papathanassiou D, Crivello F, Etard O, Delcroix N, Mazoyer B, 1036 Joliot M (2002) Automated anatomical labeling of activations in SPM using a macroscopic 1037 anatomical parcellation of the MNI MRI single-subject brain. Neuroimage 15:273-289. 1038

Uddin LQ, Supekar K, Lynch CJ, Cheng KM, Odriozola P, Barth ME, Phillips J, Feinstein C, 1039 Abrams DA, Menon V (2015) Brain State Differentiation and Behavioral Inflexibility in 1040 Autism. Cereb Cortex 25:4740-4747. 1041

Via E, Radua J, Cardoner N, Happe F, Mataix-Cols D (2011) Meta-analysis of gray matter 1042 abnormalities in autism spectrum disorder: should Asperger disorder be subsumed under a 1043 broader umbrella of autistic spectrum disorder? Arch Gen Psychiatry 68:409-418. 1044

Watanabe T, Rees G (2016) Anatomical imbalance between cortical networks in autism. Sci Rep 1045 6:31114. 1046

of 38

Watanabe T, Rees G (2017) Brain network dynamics in high-functioning individuals with autism. 1047 Nat Commun 8:16048. 1048

Watanabe T, Masuda N, Megumi F, Kanai R, Rees G (2014a) Energy landscape and dynamics of 1049 brain activity during human bistable perception. Nat Commun 5:4765. 1050

Watanabe T, Kuroda M, Kuwabara H, Aoki Y, Iwashiro N, Tatsunobu N, Takao H, Nippashi Y, 1051 Kawakubo Y, Kunimatsu A, Kasai K, Yamasue H (2015a) Clinical and neural effects of six-1052 week administration of oxytocin on core symptoms of autism. Brain 138:3400-3412. 1053

Watanabe T, Hanajima R, Shirota Y, Tsutsumi R, Shimizu T, Hayashi T, Terao Y, Ugawa Y, 1054 Katsura M, Kunimatsu A, Ohtomo K, Hirose S, Miyashita Y, Konishi S (2015b) Effects of 1055 rTMS of pre-supplementary motor area on fronto basal ganglia network activity during stop-1056 signal task. J Neurosci 35:4813-4823. 1057

Watanabe T, Yahata N, Kawakubo Y, Inoue H, Takano Y, Iwashiro N, Natsubori T, Takao H, Sasaki 1058 H, Gonoi W, Murakami M, Katsura M, Kunimatsu A, Abe O, Kasai K, Yamasue H (2014b) 1059 Network structure underlying resolution of conflicting non-verbal and verbal social 1060 information. Soc Cogn Affect Neurosci 9:767-775. 1061

Yerys BE, Antezana L, Weinblatt R, Jankowski KF, Strang J, Vaidya CJ, Schultz RT, Gaillard WD, 1062 Kenworthy L (2015) Neural Correlates of Set-Shifting in Children With Autism. Autism Res 1063 8:386-397. 1064

Yu KK, Cheung C, Chua SE, McAlonan GM (2011) Can Asperger syndrome be distinguished from 1065 autism? An anatomic likelihood meta-analysis of MRI studies. J Psychiatry Neurosci 36:412-1066 421. 1067

1068

a neuroanatomical substrate linking perceptual stability ... · page 1 of 38 í î a...

Documents